When Speed Was King: Vinod Dham and the Birth of the PentiumPublished: March 20, 2008 in India Knowledge@Wharton
It seems like only yesterday that people were tapping away on computers powered by chips called 286, 386, and so on -- the so-called "X86" series of microprocessors. Most users didn't quite understand what the chips were or did, or who actually made them -- they just had the vague understanding that the higher the number, the faster their machine would run. All of a sudden, in 1993, Intel changed the game with its launch of the first named chip -- the Pentium. Consumers had been told just a few years before to look for "Intel Inside"; now they had a name to go with it.
In an interview with India Knowledge@Wharton, Vinod Dham, widely regarded as the "Father of the Pentium" for his work as vice president and general manager of the Microprocessor Products group at Intel, shares his thoughts on those heady days of ever-faster microprocessor creation, the future of the desktop computer, and what it was really like to work with former Intel CEO Andy Grove. An edited version of the conversation appears below. Dham, who is now cofounder and managing director of the venture funds New Path Venture and NEA-IndoUS Ventures, spoke at the Wharton India Economic Forum on March 22 in Philadelphia.
India Knowledge@Wharton: Tell us about your background before you came to the United States.
Dham: I got a bachelor's degree in electrical engineering from the Delhi College of Engineering. Then I started working in the only semiconductor startup that existed in India back then. There was some activity happening in a government-sponsored division in Bangalore, where they were doing some work in semiconductors with Bharat Electronics, but in private industry there was only one company, called Continental Devices India, which was collaborating with Teledyne Semiconductor in California. I joined them and helped them put together a facility in Delhi -- I worked there for four years. My love for semiconductors started then. It continued on through my coming to the U.S., getting a [master's] degree in solid state sciences to learn more about semiconductors, and leading up to [joining] NCR, where I worked on some very advanced technologies. For several years, I worked on the technology that resulted in the invention of what we call "Flash" technology today. I brought that [experience] with me to Intel, where I spent 16 years. I moved on to other roles inside Intel, including leading microprocessor development for 386 microprocessors, 486 microprocessors, and finally the Pentium.
India Knowledge@Wharton: You are of course well known for the work that you did on the Pentium processor. What challenges did you encounter during its development?
Dham: With any microprocessor, it's very challenging to put together the design team and create an architecture that gives you a significant performance boost while at the same time meeting the schedule and getting it out on time in a cost-effective way. You have to make sure your design doesn't become so large that the product does not meet the cost target. So those challenges are there for any advanced microprocessor design, even today. But what was uniquely challenging for the Pentium program was the fact that back then, the microprocessor industry was going through a major shakeup. There were multiple architectures. There was the Intel architecture, characterized as X86. There was the SPARC architecture -- which still exists -- from Sun Microsystems, that they were primarily using for high-end workstations and servers, as opposed to the desktop or laptop products we were dominant in. Then there was an architecture that IBM had come up with. They called it PowerPC, and they were supplying this product to Apple Computer. All of the Macintosh products were based on this particular architecture.
And then right in the middle of it, around that time, there was this whole phenomenon called RISC, which stood for Reduced Instruction Set Computing, as opposed to CISC, which stood for Complex Instruction Set Computing, which is what we were using at Intel. Now, the reason we had CISC architecture was that it's how we had started our 4004 microprocessors, in the 1970s, and that had led to 8008, 8080, 286, 386 and 486. Part of the architecture was maintained in order to maintain backward compatibility, to ensure that software developed for previous-generation platforms could also run on the new microprocessors.
The RISC computing really took off in a big way with the advent of MIPS, which was the other company that came on the scene. MIPS stood for Millions of Instructions Per Second, and that was the name of the company itself -- it's still around. They said that they could build a RISC-based processor that would be much more cost effective and much higher performance than Intel's technology could build. Therefore a whole host of companies, ranging from Microsoft to Compaq, joined with MIPS to create a new platform that was to threaten the very existence of Intel, by threatening whether the Pentium would ever come to life and ever be accepted in the marketplace. So with that in the background, the development of the Pentium really was a much more challenging task than normally would be the case.
India Knowledge@Wharton: The major issue with the Pentium, which was publicized quite a bit, was the floating-point flaw. Do you recall when you first heard about the problem? What was your first reaction?
Dham: In any microprocessor or large operating system, it's not unusual to find minor bugs or issues. They come with a whole long list of errata. What was different about the Pentium was that it was the first time in the history of a semiconductor component that the product had a name, as opposed to a number. Most people don't know what a semiconductor is or what a chip is -- they've never seen one. They don't know what's inside their desktop or laptop or cellphone or iPod. Through our branding, "Pentium" had become a household word. Therefore, there was an expectation that Intel was a consumer company, not a chip company. If the consumer wasn't happy about something, the company had to act in a certain way to take care of it. Now we heard of this problem and a few months later we actually began to offer to customers that if they were not happy with the fact that under certain conditions, this component could miscalculate, we were willing to have them send that component back. We'd replace it with a good component, because by then we had started producing the correct component and we were in a position to replace it.
India Knowledge@Wharton: What did you personally and Intel as a company learn from this experience? If you found yourself in a similar situation today, would you handle it differently?
Dham: What Intel learned is that when you create a brand, you have to stand up and be accountable for it when people are not happy with the product you're offering. You have to correct the situation immediately and unconditionally, without asking any questions. You can't go around and make suggestions about "what if this is fixed this way," or "this may not harm you." All of that is moot. When you're dealing with consumers, you have to satisfy them and delight them. One way is just to give them the option upfront to return the component, and do what is necessary to keep them happy. Intel also recognized that it was no longer a semiconductor manufacturer but also a company that had now arrived on the world stage. And the world expected it to deliver a consumer product, even though Intel didn't build the PCs or the desktops or laptops but rather the component that goes inside them. Since Intel said, "Intel Inside," Intel had the obligation to take care of it -- not just Compaq or IBM or any of the companies that sold the computer. A very dramatic change had occurred -- the idea that not just the supplier, but the supplier who supplied to the supplier, also had responsibility for it.
India Knowledge@Wharton: The "Intel Inside" campaign allowed the company to evolve from a behind-the-scenes manufacturer of OEM components to a well-known consumer brand. Are there any lessons for companies in that?
Dham: When you find out there is a problem, you should immediately go out and be the first one to disclose it. Don't take your time to fix the problem, no matter what it is. Yes, the problem was rare, and errors would happen under very rare circumstances, but all that doesn't matter. You have to go out and tell the world -- "There is a problem, we are fixing it, it will take a certain amount of time, and meanwhile, if you want to continue using the machine, you are most welcome to because it's not likely to cause you to burn or crash. And, if you don't feel comfortable, feel free to send it back to us; when we are ready with our corrected part, we will send you a replacement." That would have been the most appropriate and correct response, in my mind. And that's what I would recommend others do. That's what I would do the next time, if I were in such a position again.
India Knowledge@Wharton: You worked closely with Andy Grove at Intel. What was it like to work with him?
Dham: He's a brilliant man -- a man of tremendous intellect. He came [to the U.S.] as an immigrant, and went to [City College] in New York and [then] to Berkeley. Semiconductors were his first love, but he actually went on to become a businessman and a business strategist. He reinvented himself more successfully, I think, than most people I know. He was very intense, very focused, very disciplined, very results-oriented.... Those were some of his characteristics. And he was also a person who cared enough about the company that he would carefully orchestrate every move to ensure that he got the results he wanted.
India Knowledge@Wharton: What were the best and worst things about working with him?
Dham: The best thing was the fact that you learned so much about how to develop business strategy and how to win in the marketplace -- how to be paranoid about everything that can go wrong and therefore to go the extra mile to make sure you have covered all the bases. And discipline -- to focus on what the objectives are, what the goals are, and then execution -- to deliver on those goals, monitoring them on a regular basis in a quantifiable way to ensure that you are making progress. [He had a] no-nonsense, straight-shooting style of management and culture where people were not blamed for saying the wrong thing as long as they were doing the right thing.
One negative about working with Grove was that sometimes he was too intense and overlooked the emotions of the people he was dealing with, and how they felt about whether they were right or wrong. He had a way of conveying his point of view -- sometimes he used a hammer instead of a small tap. I think that scared a lot of people who were working with him closely.
India Knowledge@Wharton: When did you leave Intel, and why?
Dham: When I joined Intel in 1979, at that time basically I was still an engineer. I moved gradually into the management ranks, leading programs and projects and running businesses, being part of upper management as a vice president, and then leading the Pentium project -- which was the greatest success Intel had in any product to date. You know, if you are here in Silicon Valley 10 to 15 years and you have not stepped out and done a startup, there's something wrong with you. I was 45 years old in 1995, and I had this major midlife crisis that I had to resolve. Either I could continue to be a tenured professor inside Intel and spend the rest of my life doing whatever management wanted and just retire at the end in a handsome way or -- I could take this plunge into the unknown dark side, of startups, and discover what it was like. I think there were enough stories of startups and great breakthroughs in those startups -- and I was financially reasonably well off -- that I felt I was ready, and that was the key reason I left.
India Knowledge@Wharton: What have you been doing with startups since you left?
Dham: I played it very safe. Coming from a big corporation, I didn't want to plunge into something totally unknown. I joined a company called NexGen, which was about eight or nine years old. They had been supposedly developing an X86 product, and I thought I could go and bring my experience from Intel and help them.
I went into NexGen as the chief operating officer and soon I discovered the chaotic and scrappy nature of working in such a company and how disillusioning it was -- sometimes you really have to be, not just a maverick, but also a little crazy to be doing startups. With my background I was very quickly able to assess what the future of this company would be, and it looked bleak. The way they were doing the technology -- basically building a very proprietary product across the board -- they would have had to build a proprietary personal computer to sell that product. That would be very hard, because the personal computer industry would frown upon building a special motherboard with special chipsets and technologies just to accommodate their processors.
I quickly had them redirect their effort to put a Pentium bus on the program. That was a breakthrough that ultimately caused the company to get bought by Advanced Micro Devices, Intel's archenemy, because they had been developing a [similar] product internally, except that it had misfired. So they basically acquired NexGen to fill that hole.
India Knowledge@Wharton: You ended up working for AMD, right, and helped develop the K6 processor? What was it like to compete with Intel? What innovations did the K6 introduce?
Dham: For a few months, [AMD's] K6 was the highest performance X86 processor in the world. That got a lot of people excited, both inside and outside Intel. We were able to create what I call a "Pentium killer." It didn't last for too long, because Intel with its might was very quickly able to assemble processors that went beyond the K6. But there was a brief moment of celebration as we created a product that was faster than what Intel was selling.
But the biggest thing was not that we built something faster -- that's not really a sustainable advantage. I think the biggest contribution I made was in creating for the first time a legitimate sub-$1,000 PC industry. Today, of course, people take it for granted. You can go out and buy a desktop for $300 or $400. But back in the 1980s and 1990s when I was at Intel, our plan always was to let the price point of the PC stay around $2,000 and continue to deliver into it a higher-performance microprocessor every year -- that is, to give higher value for the same money to the customer. There was really no intention of selling computers that were cheaper -- just selling computers at a high price point where everybody makes a lot of money, and continuing to deliver higher performance, more value and more applications on that computer.
What I created with the K6 at AMD was a sub-$1,000 machine. It forced Intel to quickly come up with a makeshift solution to counter us because they were caught totally unprepared for that type of onslaught.
India Knowledge@Wharton: Recently, there has been a huge increase in portable and mobile devices. How is that affecting the microprocessor industry?
Dham: What's happening is very interesting. Companies like Intel and AMD and others truly do have a big challenge in front of them. Back in the 1980s and 1990s, which I call the Golden Era, there was a huge demand for these machines, and the software just wouldn't keep up with the hardware. Every time we came up with a higher-performance processor, people had to rewrite software to take advantage of many more functions and to make things run faster. But now we've reached the point where most people don't even know how fast the chip inside their computer is. Most computers run fast enough that it's no longer the speed at which the chip runs but all the other functionality -- accessing the Internet, connectivity and things of that nature -- that is significantly more important.
In some ways, the industry has become very mature. IBM no longer builds the computers -- they sold the PC division to Lenovo; Compaq got bought by Hewlett-Packard. The industry is now consolidated into three or four players who build these machines. Comdex, which was the flagship show I went to for 10 years, where we used to launch all the computers and chips -- has died. There used to be tens of thousands of people gathering in Las Vegas every year celebrating, and there is no longer a need to do that. New computers come and go and people don't notice. It's becoming more of a replacement cycle for mature economies like the U.S. and Europe. People have multiple machines in their homes, and they only buy the next one if there's a compelling need -- and there's not much of a compelling need. People are sitting on one machine for four, five or six years, unlike in the 1980s or 1990s, where they had to change their computer every one to two years because the previous one just didn't run fast enough. That's one place where this industry is facing a challenge in terms of growing the overall market. Yes, more desktops are moving to laptops, so maybe you can make money by doing that, but the overall market in the advanced part of the world is not growing.
Something more interesting is happening in the developing nations now. China clearly has a leg up on India in terms of its broadband capability and connectivity, and therefore it has more computer penetration than India, but nowhere near what we have in the U.S. However, in India there is no broadband penetration. So even if you have a computer there is no way to hook it up on the Internet fast enough to allow you to share videos and share music and do all the things we normally do here [in the U.S.] with our laptops and desktops.
Most consumers in India are not using computers due to lack of broadband and are actually in love with their cell phones. Cell phones are becoming significantly smarter -- as you can see with what the iPhone has done -- and they're becoming like mini-computers. India already demonstrated successfully that it can go from not having phones at all directly to cell phones -- for example there are only 44 million landline phones in India, and there are 239 million cell phones. They've leapfrogged the landline business completely. People don't buy landline phones in India because they need not do so. Now there's a fear that Indian consumers will do the same thing to the desktop-laptop industry -- they will leapfrog it completely and instead buy smart phones. I think companies like Intel have a big challenge in front of them.